Method for image segmentation to identify regions with constant foreground color

• US 7,376,272 B2
• Filed: 06/14/2004
• Issued: 05/20/2008
• Est. Priority Date: 06/14/2004
• Status: Expired due to Fees

• Current Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Original Assignee: Xerox Corporation (Xerox Holdings Corp.)
• Inventors: Fan, Zhigang, Jacobs, Timothy W.
• Primary Examiner(s): Bella; Matthew C.
• Assistant Examiner(s): Bayat; Ali
• Application Number:
  US10/866,850
• Publication Number:
  US 20050275897A1
• Time in Patent Office: 1,436 Days
• Field of Search: 382/162, 382/164, 382/165, 382/166, 382/167, 382/173, 382/176, 382/180, 382/190, 382/232, 382/255, 382/274, 358/1.9, 358/464, 358/462, 358/2.1, 345/636
• US Class Current: 382/173
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/162 Quantising the image signal
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables
• CPC Class Codes: H04N 1/40062 Discrimination between different image types, e.g. two-tone, continuous tone

Abstract

The present invention is a method for image segmentation to produce a mixed raster content (MRC) image with constant foreground layers. The invention extracts uniform text and other uniform color objects that carry detail information. The method includes four primary steps. First, the objects are extracted from the image. Next, the objects are tested for color consistency and other features to decide if they should be chosen for coding to the MRC foreground layers. The objects that are chosen are then clustered in color space. The image is finally segmented such that each foreground layer codes the objects from the same color cluster.

Claims

1. A method for the segmentation of a digital image for representation in a mixed raster content form with a constant foreground, comprising the steps of:

extracting uniform color objects from the image;

testing at least some of the extracted objects for color consistency to decide if the extracted objects should be coded to a foreground layer in the mixed raster content form;

clustering, in color space, objects that are chosen for representation the foreground layer to associate objects in at least one common color cluster; and

segmenting the image such that each foreground layer represents objects from the common color cluster.

2. The method of claim 1, wherein said step of extracting uniform color objects includes extracting text and other objects that carry detail information.

3. The method of claim 1, wherein said step of testing for color consistency, further includes testing for other consistent features.

4. The method of claim 1, wherein said step of clustering objects further comprises testing the color of only an interior portion of the object to determine whether a first extracted object should be clustered with a second extracted object.

5. The method of claim 4, wherein said testing step comprises a thresholding operation that compares colors of objects, wherein a color difference is determined between extracted objects being compared and a threshold is then applied to the color difference to determine if the objects should be clustered.

6. The method of claim 5, wherein the threshold applied to the color difference is a function of the size of the object, such that objects of smaller size may have a greater acceptable color difference for the cluster than objects of a larger size.

7. The method of claim 5, wherein the threshold applied to the color difference is a function of the difference between the colors of the clusters being considered, such that a greater difference between cluster colors will result in a larger threshold for color differences for extracted objects within a color.

8. The method of claim 4, wherein said testing step comprises a comparison of the position of an extracted object relative to other objects within the cluster for which the object is being considered.

9. The method of claim 8, wherein said comparison of position takes into account a relationship between text characters within a line and on other lines such that the characters within a line will have a first threshold for color differences and the characters on other lines will have a second threshold for color differences that is lower than the first threshold.

10. The method of claim 1, wherein the clustering step limits the number of colors to which objects are clustered to a subset of colors in the original image.

11. The method of claim 1, wherein a color representing clustered image objects is an average of colors of objects in the cluster.

12. The method of claim 1, wherein the number of clusters is limited to a predefined number of clusters.

13. The method of claim 1, further comprising the step of altering, in response to a specified preference, at least one color for a segment of the image.

14. The method of claim 13, wherein the at least one color is a color near black and where the at least one color is altered to equal black.

15. The method of claim 14, wherein the segment is a text segment and the color for all text within the segment is set to black.